Smoke tests fail on devices in the lab

[Saviq]

Version

v2.26.0

Environment

N/A

What happened? What did you expect to happen?

Smoke tests pass

Reproduction steps

No response

Relevant log output

I: The following clients failed to execute successfully:
I:     eglinfo/virtual
I:     eglinfo/gbm-kms
I:     /snap/mir-test-tools/25106/usr/bin/mir_demo_client_mir_shell/gbm-kms
I:     /snap/mir-test-tools/25106/usr/bin/mir_demo_client_wayland/gbm-kms
I:     /snap/mir-test-tools/25106/usr/bin/mir_demo_client_wayland_egl_spinner/gbm-kms
I:     eglinfo/atomic-kms
I:     /snap/mir-test-tools/25106/usr/bin/mir_demo_client_mir_shell/atomic-kms
I:     /snap/mir-test-tools/25106/usr/bin/mir_demo_client_wayland/atomic-kms
I:     /snap/mir-test-tools/25106/usr/bin/mir_demo_client_wayland_egl_spinner/atomic-kms
I:     eglinfo/eglstream-kms
I:     /snap/mir-test-tools/25106/usr/bin/mir_demo_client_mir_shell/eglstream-kms
I:     /snap/mir-test-tools/25106/usr/bin/mir_demo_client_wayland/eglstream-kms
I:     /snap/mir-test-tools/25106/usr/bin/mir_demo_client_wayland_egl_spinner/eglstream-kms
I: Smoke testing complete with returncode -1, platforms tested: virtual gbm-kms atomic-kms eglstream-kms

[Saviq]

This is (partially) fallout from executing all the platforms in a sequence.

[AlanGriffiths]

This is (partially) fallout from executing all the platforms in a sequence.

If each of the platforms works, why would executing in a sequence be expected to cause problem?

[Saviq]

If each of the platforms works, why would executing in a sequence be expected to cause problem?

I'm not saying it's the cause, but it is a fallout. At least part of it is that we're running them in quick succession, and we're not cleaning up properly - if one hangs on exit, all subsequent ones fail with DRM permission errors trying to grab DRM master.

[Saviq]

Another thing is that we stopped testing the default case.

And on hybrid systems some platforms might not have displays connected…